Super Regulator, collecting the facts - Page 25 - diyAudio
Go Back   Home > Forums > Amplifiers > Power Supplies

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 30th October 2003, 03:51 AM   #241
jwb is offline jwb  United States
diyAudio Member
 
jwb's Avatar
 
Join Date: Mar 2002
Location: San Francisco, USA
Send a message via AIM to jwb
Quite right.
  Reply With Quote
Old 30th October 2003, 08:05 AM   #242
diyAudio Retiree
 
Join Date: Oct 2002
Location: Spain or the pueblo of Los Angeles
Thumbs up someone is paying attention.......

"Quite right" My fault for not explaining that it was already part of the regulator. Of course the gate of J202 could also be biased from the filtered output of the LM329 for less noise. If you used a film cap and a large resistor to filter the reference voltage at the op amp, you might have to watch out for the effect of 20 to 30nA of leakage current from the reverse biased diode. A one megohm resistor in this filter would only give you only a few 10s of millivolts error though, and is probably not going to worry too many people. The 5K source resistor and voltage at the gate, would give about 2 to 3 times the output impedance over the J202 without any source resistor. This would be without the extra complexity and expense of a cascode current source. I don't want to over do the circuit either, and a cascoded current source is getting in the realm of the truly exotic for this application. This current source is providing the current to bias a fairly low impedance LED and we are probably below the LED noise floor of a few microvolts for a single jfet current source with a couple of megohms output impedance. It has been very interesting to look so closely at the performance of jfet current sources and LED voltage references. I would like to thank Mr. Curl and Mr. Jung for their thorough analysis of these circuit options. Where else are you going to get stuff this good guys, without paying for pretty specialized consulting? I am pleased for the dialog on a subject that's very interesting to me, and hopefully to others. I appreciate both of these gentlemen's efforts greatly.

Thanks JC and WJ,

Fred Dieckmann
  Reply With Quote
Old 30th October 2003, 04:53 PM   #243
Dave S is offline Dave S  United Kingdom
diyAudio Member
 
Join Date: Mar 2003
Location: Searchin' for Sweet Singletrack
We seem to be convinced that the opamp supply should be derived from the regulated output, which then gets us into all the level shifting and current source discussions (very enlightning).

I would like to suggest an alternative method that avoids the level shifter (at least inside the feedback loop). I would welcome your comments on the approach shown in the attachment (apologies I have not yet learned how to insert an image into a post). R3 is probably is not required because the values of R1 and R2 can be adjusted to sink the correct current.

Having one less transistor in the f/b loop may help with stability and transient response. Also the finite current source impedance will affect only the supply to the opamp rather than a hgih impedance point within the feedback loop.
Attached Files
File Type: zip jungreg.zip (5.5 KB, 178 views)
  Reply With Quote
Old 30th October 2003, 05:19 PM   #244
jwb is offline jwb  United States
diyAudio Member
 
jwb's Avatar
 
Join Date: Mar 2002
Location: San Francisco, USA
Send a message via AIM to jwb
Same as above.
Attached Images
File Type: png jungreg.png (3.8 KB, 1185 views)
  Reply With Quote
Old 30th October 2003, 06:01 PM   #245
jcx is offline jcx  United States
diyAudio Member
 
Join Date: Feb 2003
Location: ..
hi Walt

the more fundamental question that interests me is how does the super reg compensation work, especially when many diy hackers may hook the super reg up to circuits with 1000’s of uF of local supply bypass C

I scanned the AD825 SPICE model to see if it would realistically model the 4/2000 super regulator circuit because it appears to me that the regulator either relies on the special capacitive load stabilization compensation of the AD825 or the relatively high 1st pole of the open loop gain causing the load Cap*(Requiv op amp+ Q1) to form the dominant feedback pole. Further the super reg has a minor feedback coupling through the PSRR (is the negative voltage version noticably different due to the bootstrapping of the internal compensation cap? Is the AD817 Ccomp referenced to the positive rail instead of the negative rail as is the case with the AD825 and would one be preferred over the other in pos/neg super reg? )

having looked at the AD825 SPICE models, I don't see where the capacitive load gain reduction/stabilization from Cf in the simplified schematic (ad825 datasheet fig 4) is modeled, unless reducing L1 to the physically ludicrous value of 6 pH is supposed to cover it

has anyone at Analog Devices other than Alexander and Bowers read an-138?
certainly few of the apps people you (used to?) let write SPICE models seem to have taken it for a standard to aspire to
  Reply With Quote
Old 30th October 2003, 06:31 PM   #246
WaltJ is offline WaltJ  United States
diyAudio Member
 
WaltJ's Avatar
 
Join Date: Aug 2003
Location: USA
Default Re: POST #469

Quote:
the more fundamental question that interests me is how does the super reg compensation work, especially when many diy hackers may hook the super reg up to circuits with 1000’s of uF of local supply bypass C
It depends upon a finite ESR to work. That is why high-Q caps can make it sing if nearby. ESR of 100-200milliohms or more is best, and very low ESR lytics (for example Oscons) can also give trouble, by giving rise to quasi-stability (doesn't oscillate outright, but will ring with step loads).

Quote:
I scanned the AD825 SPICE model to see if it would realistically model the 4/2000 super regulator circuit because it appears to me that the regulator either relies on the special capacitive load stabilization compensation of the AD825 or the relatively high 1st pole of the open loop gain causing the load Cap*(Requiv op amp+ Q1) to form the dominant feedback pole.
The regulator is not dependent upon the ad825 or ad817 internal cap load compensation for operation. Many regs have been built with other op amps. That said, I would also say that pole/zero compensated op amps may give trouble, because of their phase funnies (phase margin dips) at HF. See OP27, for example.

Quote:
having looked at the AD825 SPICE models, I don't see where the capacitive load gain reduction/stabilization from Cf in the simplified schematic (ad825 datasheet fig 4) is modeled, unless reducing L1 to the physically ludicrous value of 6 pH is supposed to cover it
I don't believe this part of the ad825's circuit made it to the model, at least it isn't obvious to me that it did. I wasn't involved with the ADI model generation, but I am familiar with the Derek Bowers/Mark Alexander topology, which the ad825 does use. I will forward your comments to the apps folks there at ADI.

wj
  Reply With Quote
Old 31st October 2003, 03:35 PM   #247
jcx is offline jcx  United States
diyAudio Member
 
Join Date: Feb 2003
Location: ..
I really think the 4/2000 super reg with the AD825 is operating with the output cap and Q1 drive impedance forming the dominant pole;

[[~ 8 Ohm (AD825 Zout) + 10 Ohm (R23) + Rb (Q1) ] / ~100 (hfe Q1) ]+ Re (Q1)

~>/= 0.3 Ohms

0.3 Ohm * 120 uF ~ 30 uS

the AD825 open loop pole is:

2000 (open loop gain) * ~= 8 nS ( from 1/ 2pi*GBW) = 16 uS

the Cout time constant is 2 x the AD825 open loop pole

I wouldn’t consider this approach to have good margin due to with the likely large variability of Hfe and AD825 open loop gain, but increasing ouput Cap C4 by 8-10X is entirely practical and this compensation scheme would work with extremely low ESR caps

A Local (inner loop) feedback could stabilize the AD825 open loop gain pole location, the loop gain loss from setting inner loop to ~ 1000 is small and the reg Zout increase from reduced loop gain is easily made up by increased Cout, the inner loop could also be tailored to control possible strangeness above 1 MHz where the output cap Z is rising and Q1 Hfe roll off begins entering the picture while the AD825 still has gain left

I realize the mods to make this scheme work are well beyond the point of your 4/2000 articles’ use the existing pcb


Re multipole macro models; the AD825 is a simplified version of AN138 and fails to capture the different +/- PSRR in the more realistic way that proportioning pole/zero impedances between the op amp power pins is capable of
also isn’t Cin really fet body-substrate C and should at least be modeled with diodes to capture some of the input Z common mode input nonlinearity
  Reply With Quote
Old 31st October 2003, 04:00 PM   #248
dimitri is offline dimitri  United States
diyAudio Member
 
dimitri's Avatar
 
Join Date: Oct 2002
Location: retired
Send a message via ICQ to dimitri
Default to jcx

0.3 Ohm * 120 uF ~ 30 uS
this is the pole time constant

but there is a _zero_ time constant too
(0.3 Ohm ll ESR)* 120 uF
  Reply With Quote
Old 31st October 2003, 04:28 PM   #249
WaltJ is offline WaltJ  United States
diyAudio Member
 
WaltJ's Avatar
 
Join Date: Aug 2003
Location: USA
Default also to jcx

wj: See below for "wj: " comments...

I really think the 4/2000 super reg with the AD825 is operating with the output cap and Q1 drive impedance forming the dominant pole;

[[~ 8 Ohm (AD825 Zout) + 10 Ohm (R23) + Rb (Q1) ] / ~100 (hfe Q1) ]+ Re (Q1)

~>/= 0.3 Ohms

0.3 Ohm * 120 uF ~ 30 uS

wj: Actually the math above yields 36 us, not 30.

wj: And, there is a zero at Resr*C, which is 0.15 * 120e-6 = 18 us. This roughly falls on top the AD825 OL pole. gee... that seems really fortuitous, doesn't it?


the AD825 open loop pole is:

2000 (open loop gain) * ~= 8 nS ( from 1/ 2pi*GBW) = 16 uS

the Cout time constant is 2 x the AD825 open loop pole

I wouldn’t consider this approach to have good margin due to with the likely large variability of Hfe and AD825 open loop gain, but increasing ouput Cap C4 by 8-10X is entirely practical and this compensation scheme would work with extremely low ESR caps

wj: I don't see the logic of this, as your analysis has neglected the cap ESR altogether! This cannot be overlooked. Obviously, as I pointed out earlier it does enter into the picture, and it can influence stability if out of range. And, I actually noted the stability degradation with low ESR caps. A perfect cap would be problematic..... as was also noted before (comment on high-Q caps).

wj: Ahah! Dmitri's post #472 has just arrived, so he has beat me to pointing out your error on neglecting ESR. Thanks for the comments, Dmitri.


A Local (inner loop) feedback could stabilize the AD825 open loop gain pole location, the loop gain loss from setting inner loop to ~ 1000 is small and easily made up by increased Cout, the inner loop could also be tailored to control possible strangeness above 1 MHz where the output cap Z is rising and Q1 Hfe roll off begins entering the picture while the AD825 still has gain left

I realize the mods to make this scheme work are well beyond the point of your 4/2000 articles’ use the existing pcb


wj: Actually, a workable OL compensation *was* discussed in the article, as an option. Did you miss it?


Re multipole macro models; the AD825 is a simplified version of AN138 and fails to capture the PSSR in the more realistic way that proportioning pole/zero impedances between the op amp power pins is capable of
also isn’t Cin really fet body-substrate C and should at least be modeled with diodes to capture some of the input Z common mode input nonlinearity

wj: I don't know where you found the axe to grind here. I don't write the ADI models, and I no longer work for them. I suggest that you write to linear.apps@analog.com, if you feel so strongly about macromodeling sins of omission.


wj
  Reply With Quote
Old 31st October 2003, 05:54 PM   #250
diyAudio Retiree
 
Join Date: Oct 2002
Location: Spain or the pueblo of Los Angeles
Default Small cap gains

Mr. Jung raises the excellent point that an output capacitors parasitics sold be part of the design process. I believe this topic has come up on the forum in the past as well. The ESR is often given in the capacitor specs and ESL can be estimated by compare the case size of the cap in question with that of a similar cap who's ESL number is characterized. This level of parasitics analysis is becoming common enough that at least one manufacturer has developed a Spice program for their capacitors.

http://www.kemet.com/kemet/web/homep...name/spicesoft

Deliberately increasing the ESR of capacitor types such as OSCON Black and film caps can also be achieved with a small series resistor. A parallel array of small low ESR and ESL caps, each having it's own additional resistor could give quite a bit of flexibility in creating the characteristics one might want in this "composite capacitor"
Attached Images
File Type: jpg esrtune.jpg (25.1 KB, 979 views)
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



New To Site? Need Help?

All times are GMT. The time now is 04:02 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2